Checklist

• I have searched related issues but cannot get the expected help.
• 2. I have read the FAQ documentation but cannot get the expected help.
• 3. The bug has not been fixed in the latest version.

Describe the bug

C:\Users\ASUS\AppData\Local\Programs\Python\Python311\python.exe C:\Users\ASUS\Desktop\newPro\mmdeploy\test.py
ONNXRuntime 版本 1.16.0
Pytorch 版本 2.1.0+cu118
Pytorch 版本 0.16.0+cu118
CUDA 是否可用 True
11/18 21:16:22 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "Codebases" registry tree. As a workaround, the current "Codebases" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/18 21:16:22 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "mmseg_tasks" registry tree. As a workaround, the current "mmseg_tasks" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
c:\users\asus\desktop\newpro\mmsegmentation\mmseg\models\decode_heads\decode_head.py:120: UserWarning: For binary segmentation, we suggest usingout_channels = 1 to define the outputchannels of segmentor, and use thresholdto convert seg_logits into a predictionapplying a threshold
warnings.warn('For binary segmentation, we suggest using'
c:\users\asus\desktop\newpro\mmsegmentation\mmseg\models\losses\cross_entropy_loss.py:250: UserWarning: Default avg_non_ignore is False, if you would like to ignore the certain label and average loss over non-ignore labels, which is the same with PyTorch official cross_entropy, set avg_non_ignore=True.
warnings.warn(
Loads checkpoint by local backend from path: FPCDataset-Segformer/iter_40000.pth
11/18 21:16:23 - mmengine - WARNING - DeprecationWarning: get_onnx_config will be deprecated in the future.
11/18 21:16:23 - mmengine - INFO - Export PyTorch model to ONNX: work_dir/end2end.onnx.
C:\Users\ASUS\Desktop\newPro\mmdeploy\mmdeploy\core\optimizers\function_marker.py:160: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
ys_shape = tuple(int(s) for s in ys.shape)
c:\users\asus\desktop\newpro\mmsegmentation\mmseg\models\utils\shape_convert.py:15: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
assert L == H * W, 'The seq_len doesn't match H, W'
C:\Users\ASUS\Desktop\newPro\mmdeploy\mmdeploy\codebase\mmseg\models\segmentors\base.py:61: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
if seg_logit.shape[1] == 1:

进程已结束，退出代码为 -1073741819 (0xC0000005)

Reproduction

import onnxruntime as ort
print('ONNXRuntime 版本', ort.version)

import torch, torchvision

print('Pytorch 版本', torch.version)
print('Pytorch 版本', torchvision.version)
print('CUDA 是否可用', torch.cuda.is_available())

#python tools/deploy.py configs/mmseg/segmentation_onnxruntime_dynamic.py FPCDataset-Segformer/FPCDataset_Segformer_20251012.py FPCDataset-Segformer/iter_40000.pth FPCDataset-Segformer/_2-1_OK_2025-10-08-16-53-02-934.png --work-dir mmseg2onnx_segformer --dump-info

#python tools/deploy.py configs/mmseg/segmentation_tensorrt-int8_static-512x512.py FPCDataset-Segformer/FPCDataset_Segformer_20251012.py FPCDataset-Segformer/iter_40000.pth FPCDataset-Segformer/_2-1_OK_2025-10-08-16-53-02-934.png --work-dir mmseg2ttr_segformer --device cuda:0

from mmdeploy.apis import torch2onnx
from mmdeploy.backend.sdk.export_info import export2SDK

img = 'FPCDataset-Segformer/_2-1_OK_2025-10-08-16-53-02-934.png'
work_dir = 'work_dir/'
save_file = 'end2end.onnx'
deploy_cfg = 'configs/mmseg/segmentation_onnxruntime_dynamic.py'
model_cfg = 'FPCDataset-Segformer/FPCDataset_Segformer_20251012.py'
model_checkpoint = 'FPCDataset-Segformer/iter_40000.pth'
device = 'cpu'

1. convert model to onnx

torch2onnx(img, work_dir, save_file, deploy_cfg, model_cfg,
model_checkpoint, device)

2. extract pipeline info for sdk use (dump-info)

export2SDK(deploy_cfg, model_cfg, work_dir, pth=model_checkpoint, device=device)

from mmdeploy.apis import torch2onnx
from mmdeploy.apis.tensorrt import onnx2tensorrt
from mmdeploy.backend.sdk.export_info import export2SDK
import os

1. convert model to IR(onnx)

torch2onnx(img, work_dir, save_file, deploy_cfg, model_cfg,
model_checkpoint, device)

2. convert IR to tensorrt

onnx_model = os.path.join(work_dir, save_file)
save_file = 'end2end.engine'
model_id = 0
device = 'cuda'
onnx2tensorrt(work_dir, save_file, model_id, deploy_cfg, onnx_model, device)

3. extract pipeline info for sdk use (dump-info)

export2SDK(deploy_cfg, model_cfg, work_dir, pth=model_checkpoint, device=device)

Environment

ONNXRuntime 版本 1.16.0
Pytorch 版本 2.1.0+cu118
Pytorch 版本 0.16.0+cu118
CUDA 是否可用 True
11/18 21:21:04 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "Codebases" registry tree. As a workaround, the current "Codebases" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
11/18 21:21:04 - mmengine - WARNING - Failed to search registry with scope "mmseg" in the "mmseg_tasks" registry tree. As a workaround, the current "mmseg_tasks" registry in "mmdeploy" is used to build instance. This may cause unexpected failure when running the built modules. Please check whether "mmseg" is a correct scope, or whether the registry is initialized.
c:\users\asus\desktop\newpro\mmsegmentation\mmseg\models\decode_heads\decode_head.py:120: UserWarning: For binary segmentation, we suggest using`out_channels = 1` to define the outputchannels of segmentor, and use `threshold`to convert `seg_logits` into a predictionapplying a threshold
  warnings.warn('For binary segmentation, we suggest using'
c:\users\asus\desktop\newpro\mmsegmentation\mmseg\models\losses\cross_entropy_loss.py:250: UserWarning: Default ``avg_non_ignore`` is False, if you would like to ignore the certain label and average loss over non-ignore labels, which is the same with PyTorch official cross_entropy, set ``avg_non_ignore=True``.
  warnings.warn(
Loads checkpoint by local backend from path: FPCDataset-Segformer/iter_40000.pth
11/18 21:21:05 - mmengine - WARNING - DeprecationWarning: get_onnx_config will be deprecated in the future.
11/18 21:21:05 - mmengine - INFO - Export PyTorch model to ONNX: work_dir/end2end.onnx.
C:\Users\ASUS\Desktop\newPro\mmdeploy\mmdeploy\core\optimizers\function_marker.py:160: TracerWarning: Converting a tensor to a Python integer might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  ys_shape = tuple(int(s) for s in ys.shape)
c:\users\asus\desktop\newpro\mmsegmentation\mmseg\models\utils\shape_convert.py:15: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  assert L == H * W, 'The seq_len doesn\'t match H, W'
C:\Users\ASUS\Desktop\newPro\mmdeploy\mmdeploy\codebase\mmseg\models\segmentors\base.py:61: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
  if seg_logit.shape[1] == 1:

Error traceback